7. Probabilistic Risk Assessment - University of Glasgow - …johnson/teaching/safety/powerpoint… ·  · 2011-09-27• PRA: Probabilistic Risk Assessment. ... PRA: Military vs

Probabilistic Risk Assessment

Prof. Chris Johnson,School of Computing Science, University of [email protected]://www.dcs.gla.ac.uk/~johnson

Introduction

• PRA: Probabilistic Risk Assessment.

The use of PRA technology should be increased in all regulatory matters to the extent supported by the state of the art in PRA methods and data and in a manner that complements the NRC's deterministic approach and supports the NRC's traditional defense-in-depth philosophy.PRA and associated analyses (e.g., sensitivity studies, uncertainty analyses, and importance measures) should be used in regulatory matters, where practical within the bounds of the state of the art, to reduce unnecessary conservatism associated with current regulatory requirements, regulatory guides, license commitments, and staff practices.

NRC - REGULATORY GUIDE 1.177 An Approach for Plant-Specific, Risk-Informed Decision making: Technical Specifications

Hazard Analysis vs PRA

• FMECA - hazard analysis.

• PRA part of hazard analysis.

• What is the scope of this approach?– hardware failure rates (here)?– human error rates (here)?– software failure rates (later)?

The ‘Bathtub’ Model

Probabilistic Risk Assessment

• Mechanical systems reflect bathtub model:– bed-down failure rates;– degrade failure rates;

• Electronic systems approximate stable fault rate?

• Software fault rates spike around upgrades…

• 0.2 failures per hour:– MTTF = 1 / 0.2 = 5 hrs.

PRA - Sources of Data

• MIL-HDBK-217: – Reliability Prediction of Electronic Equipment

• Failure rate models for:– ICs, transistors, diodes, resistors,– relays, switches, connectors etc.

• Field data + simplifying assumptions.

• Continual need for revision?

PRA: Military vs Industry

• MIL-HDBK-217:– too pessimistic for companies...

• Bellcore (Telcordia):– reliability prediction procedure;– AT&T's 173 Defects-Per-Million calls (99.98%).– Business critical not safety critical.

• Commercial reliability databases.– But MTTF doesn’t consider repair;– MTTR considers observations.

Probabilistic Risk Assessment

• MIL-HDBK-338B (1,000+ pages!).

• Gives no. of failures per hour per mode.

• CR = α x β x λ– CR - criticality level;– α - failure mode frequency ratio;– β - loss probability of item from mode;– λ - base failure rate for item.

• Criticality defined subjectively in FMECA.

MIL-HDBK-338B

HRA: Human Reliability Analysis

• We focussed on hardware devices.

• PRA for human reliability?

• Probably not a good idea:– Do all people have same base error probability?– Performance Shaping Factors…– Mitigations – training, cross-checking etc.

• But for completeness...– THERP is a type of HRA…

Technique for Human Error Rate Prediction

“The THERP approach uses conventional reliability technology modified to account for greater variability and independence of human performance as compared with that of equipment performance... The procedures of THERP are similar to those employed in conventional reliability analysis, except that human task activities are substituted for equipment outputs.'' (Miller and Swain, 1987).

A.D. Swain and H.E. Guttman, Handbook of Human Reliability with Emphasis on Nuclear Power Plant Applications, NUREG-CR-1278, 1985.

Technique for Human Error Rate Prediction

•

THERP - External PSFs

Ack: A.D. Swain, Comparative Evaluation of Methods for Human Reliability Analysis, (GRS-71), Garching FRG: Gesellschaft fur Reaktorsicherheit.

THERP - Stressor PSFs

THERP - Internal PSFs

Technique for Human Error Rate Prediction

• Calculate effect of PSF on HEP• - ignores WHY they affect performance.

• Succeeds or fails on Performance Shaping Factors (PSFs).

• “Psychological vacuous” (Hollnagel).

• No model of cognition etc.

CREAM

• E. Hollnagel, Cognitive Reliability and Error Analysis Method, Elsevier, Holland, 1998.

• HRA + theoretical basis.

• Simple model of control:– scrambled - unpredictable actions;– opportunistic - react dont plan;– tactical - procedures and rules;– strategic - consider full context.

CREAM - Simple Model of Control

CREAM - Simple Model of Control

CREAM

• Much more to the technique...

• But in the end...– Strategic = 0.000005 < p < 0.01– Tactic = 0.001< p < 0.1– Opportunistic = 0.01 < p < 0.5– Scrambled = 0.1 < p < 1.0

• Common performance conditions to:– probable control mode then to– reliability estimate from literature.

Conclusions

• PRA for hardware:– widely accepted with good data.

• PRA for human performance:– - many are skeptical;– - THERP -> CREAM -> ???

• PRA for software?– Will cover this soon…

Any Questions…